Solution: go back to using -Wl,--version-script.
Use ax_check_vscript.m4 from the autoconf-archive to detect support on
multiple platforms (eg Solaris ld(1) -M).
libtool -export-symbols-regexp used ld(1) --retain-symbols-file under
the hood, the latter lets some C++ weak symbols make their way into the
dynamic symbols table, along with the zmq_* interface. The reason for
such behavior is unknown to me.
Solution:
This is an issue with the imported Autoconf M4 macro package for standardised
code coverage builds, i.e. using --enable-code-coverage.
The simplest way that I could find is to add a case statement that checks if
the output of running `gcov -version` contains the "LLVM" keyword; if that is
true then do not link with LIBGCOV as its neither required nor supported when
using the GCOV frontend for LLVM; least not on Mac OS X. The case statement
would also be the most portable.
Moreover, using the "-version" argument instead of "-v" seems to be the best
bet as that is supported by the normal GCOV and LLVM GCOV frontend.
Upstream candidate - this solution should be improved by Autoconf M4 macro
overlords and applied to the upstream M4 package; I could not find a suitable
way to detect if LLVM GCOV is being used, except for the solution herein; this
should also work on *BSD too.
Solution: import ax_valgrind_check.m4 macro file to provide a
conveniente automake hook to run Valgrind on all tests.
Add --enable-valgrind to ./configure call and then run make
check-valgrind to run memcheck, helgrind, drd and sgcheck on all
tests. Run check-valgrind-memcheck to run only memcheck.
Solution: import ax_code_coverage.m4 from autoconf-archive and use it
in configure.ac and Makefile.am in order to provide a make
check-code-coverage target behind a --enable-code-coverage configure
flag, that can be used to generate a gcov/lcov code coverage report.
Depends on having gcov and lcov installed.
